Reliability and validity help you evaluate Geography fieldwork in different ways. Reliability concerns the consistency and dependability of the evidence and method. Validity concerns whether the investigation and conclusion address what they are intended to address. Consistent answers to the wrong question still leave a problem.
The quickest useful reasoning is to name the weakness, explain how it affects the finding and choose an improvement that addresses that weakness. More observations can help with some problems; clearer wording, matched timing or a different measure may be needed for others.
This guide supports geographical-methods learning in the 2027 standalone G2 Geography syllabus K226. It develops evaluation through invented examples, rather than supplying a universal paragraph to memorise. Use fieldwork and sampling for collection design and questionnaire design when wording is the main issue.
eduKateSG · Geography learning guide
Find your next learning step
Choose the question closest to your current task. The full teaching chapters stay open below, and the grouped index helps you return to a particular explanation.
Choose a chapter
Chapters 1–7 · Separate the two questions
- Ask how dependable the evidence is
- Ask whether the evidence answers the intended question
- The two ideas can identify different weaknesses
- Trace question, measure, finding and conclusion
- Accuracy and consistency are related but distinct
- Random variation and systematic problems need different treatment
- Different times can complicate a spatial comparison
Chapters 8–14 · Diagnose collection problems
- Observation duration affects comparability
- Instrument position defines the measurement context
- Different observers need a shared recording rule
- Keep original values and identifiers together
- Question wording affects what a response means
- A pilot tests the method before full collection
- Sample coverage determines which conditions are represented
Chapters 15–21 · Connect samples and conclusions
- More observations help when they address the uncertainty
- Non-response can change whose views are recorded
- Repeated measurements need comparable conditions
- An average summarises evidence without repairing every weakness
- Different methods can answer connected parts of a question
- A reliable pattern does not automatically prove cause
- A conclusion can be useful within a narrower boundary
Chapters 22–28 · Choose targeted improvements
- Worked task: choose an improvement for a timing problem
- Worked task: repair a mismatch between aim and measure
- Worked task: interpret a short repeated-count series
- Write evaluation as problem, consequence and improvement
- An unusual reading needs a reasoned decision
- Instrument resolution limits the differences you can interpret
- An unexpected result can still be a successful investigation
Chapters 29–31 · Practise a reasoned evaluation
Reliability concerns whether the method and evidence are sufficiently consistent and dependable for the intended use. Repeated or comparable observations can help assess this, but consistency must be interpreted in context.
If conditions genuinely change, different readings can be correct. Rainfall, pedestrian flow and temperature vary. The task is to distinguish that real variation from inconsistency introduced by collection or recording.
An invented count repeated at the same site on different days may differ because use changes. That does not automatically prove the method unreliable.
Describe the collection conditions and the source of uncertainty. A useful evaluation explains why a reading may be unstable or a comparison difficult. It should not treat every difference as error or every repeated number as proof of quality.
CHAPTER 2 OF 31 · Separate the two questions
2. Ask whether the evidence answers the intended question
Validity concerns the relationship between the investigation’s purpose and what its evidence can support. A clear aim and suitable measure are central.
An invented study asks whether a park supports biodiversity but collects only visitor satisfaction ratings. Those ratings can describe experience; they do not directly establish the biodiversity claim.
The questionnaire may be delivered consistently, yet the measure remains poorly matched to the original question.
A targeted improvement might add appropriate ecological evidence or narrow the aim to visitor experience. Either route should be explained. Validity is not a label attached to a neat form. It requires a clear connection among question, method, evidence and conclusion. This connection gives the final finding meaning.
CHAPTER 3 OF 31 · Separate the two questions
3. The two ideas can identify different weaknesses
A method can produce consistent readings while measuring an unsuitable variable. Another can target a relevant variable but collect it inconsistently. The problems differ and need different responses.
Use an invented example. Repeatedly measuring a wall’s surface temperature does not answer a question about ambient air temperature merely because the readings agree closely.
Moving to an appropriate air-measurement method addresses the variable mismatch. Consistent positioning and timing then support a more dependable comparison.
This is why a good evaluation avoids choosing an improvement from a memorised list before diagnosing the problem. Ask what the investigation needs to know and what the method actually records. The answer often makes the useful improvement visible without requiring a complicated checklist.
CHAPTER 4 OF 31 · Separate the two questions
4. Trace question, measure, finding and conclusion
Write the research question, measured variable, recorded finding and final claim as separate statements. Then examine the links.
An invented question concerns how park use varies through the week. A single midday count on Tuesday provides one observation, while a conclusion about the whole week exceeds that coverage.
The count may be accurate for its interval. The weakness lies in the broader inference and sampling design rather than necessarily in the counting technique.
A useful improvement adds relevant times or narrows the conclusion. This preserves the value of the original observation while addressing the unsupported extension. Evaluation becomes more precise when the location of the problem in the chain is clear.
CHAPTER 5 OF 31 · Separate the two questions
5. Accuracy and consistency are related but distinct
A method can give repeated similar values that are systematically displaced from an appropriate reference. Consistency alone does not establish accuracy.
Imagine an invented instrument that reads two units high on every check. Repetition reproduces the same bias. Averaging those readings does not automatically remove it.
Checking the instrument against a suitable reference or following an appropriate calibration procedure addresses a different problem from collecting more repetitions.
For a school task, use the equipment instructions and evidence provided. Do not invent a calibration fault merely because results are surprising. The point is to identify the possible mechanism and the information needed to assess it. A precise diagnosis leads to an improvement with a clear purpose.
CHAPTER 6 OF 31 · Separate the two questions
6. Random variation and systematic problems need different treatment
Repeated observations can help reveal variation and reduce the influence of some random fluctuations in a suitable summary. A systematic problem may persist across every repetition.
An invented counting error that sometimes misses a person differs from always omitting one entrance from the study area. More counts at the same entrance may help with the first issue but leave the second untouched.
Name the problem’s pathway. Is the method inconsistently recording the chosen observation, or consistently excluding a relevant part of the setting?
The distinction should guide the improvement. Repetition, revised coverage and instrument checks are useful for different reasons. A strong answer explains those reasons rather than attach the word reliable to any method that sounds more elaborate.
CHAPTER 7 OF 31 · Separate the two questions
7. Different times can complicate a spatial comparison
If one site is measured in the morning and another in the afternoon, the difference may reflect both location and changing conditions. This affects an attempt to isolate a spatial relationship.
An invented cooler garden reading at 8 am and warmer roadside reading at 2 pm cannot cleanly demonstrate a location-only effect.
Matched or simultaneous timing can reduce that confounding influence where feasible. Repeated comparable collection can help establish whether a pattern persists.
The improvement is specific: it addresses time differences in a location comparison. Better equipment alone would not solve that design problem. State the limitation, its consequence and the targeted revision so the reader can see why the method would improve.
CHAPTER 8 OF 31 · Diagnose collection problems
8. Observation duration affects comparability
Counts over different durations should not be compared as though the opportunity for observations were equal. A twenty-minute count naturally has a different exposure time from a five-minute count.
An invented 40 people in five minutes gives an average rate of eight per minute, while 60 in twenty gives three per minute. Raw totals and rates tell different stories.
Standardising duration or calculating an appropriate rate can support comparison. However, a rate still summarises the chosen interval and may conceal short peaks.
Explain both the benefit and remaining scope. The revision improves one aspect of comparability; it does not make every day, site and condition equivalent. Good evaluation recognises what the improvement actually changes.
CHAPTER 9 OF 31 · Diagnose collection problems
9. Instrument position defines the measurement context
Position can affect what is measured. A temperature sensor against a hot surface, a wind observation behind a building or a sound reading beside an unusual source may not represent the intended ambient condition.
Use the investigation’s aim to assess relevance. The hot surface could be exactly what a surface-temperature study wants, while unsuitable for an air-temperature comparison.
Follow the school’s approved procedures and equipment guidance. Record relevant conditions and use comparable positioning.
The evaluation should connect position to the variable, rather than declare one location universally wrong. A measure is appropriate through its relationship to the question. This makes the reasoning more useful than a generic instruction to place equipment in a better spot.
CHAPTER 10 OF 31 · Diagnose collection problems
10. Different observers need a shared recording rule
Observer differences can affect counts or classifications. One person may count only people entering; another may count everyone passing nearby. The totals then represent different events.
Agree on the observation unit and rule before collection. A short practice comparison can reveal disagreement while changes are still easy to make.
The purpose is consistency in the intended measure, not forcing observers to produce identical totals despite different real conditions.
An improvement might define the boundary, direction and interval clearly, then check shared understanding. Explain how that revision affects comparability. It will not automatically broaden the sample or establish causation, but it can repair the specific recording inconsistency that made the original results difficult to interpret.
CHAPTER 11 OF 31 · Diagnose collection problems
11. Keep original values and identifiers together
Recording errors can break the connection between a reading and its site or time. A value without context may be difficult to interpret even if it was measured accurately.
Use clear rows or fields for the observation identifier, variable, unit and relevant conditions. Check entries promptly when the original context is available.
An invented table pairing a distance from one site with a count from another creates false evidence. Sorting complete rows preserves the pairs.
The scatter-graph guide shows why this matters. A graph can look convincing while representing corrupted pairs. Evaluation should therefore examine the data handling as well as the field technique.
CHAPTER 12 OF 31 · Diagnose collection problems
12. Question wording affects what a response means
A questionnaire item combining cleanliness and access leaves a respondent with an ambiguous task if they judge the two differently. Consistent delivery does not remove that ambiguity.
Separate the dimensions when both matter and use suitable response options. The resulting answers can be interpreted more specifically.
Leading wording creates another issue by suggesting the preferred answer. Neutral phrasing allows respondents to report different experiences.
Use the questionnaire-design guide for detailed examples. Here, the evaluation task is to explain how wording affects the evidence and why the revision helps. Increasing sample size would reproduce the same flawed item more widely. A targeted change repairs the interpretation problem at its source.
CHAPTER 13 OF 31 · Diagnose collection problems
13. A pilot tests the method before full collection
A small pilot can reveal unclear instructions, unsuitable categories or practical recording difficulties. Its value comes from what is learned and changed.
Ask how participants understood an item or compare how observers applied a rule. A pilot producing a tidy pattern does not by itself establish that the method works as intended.
Revise the identified issue and test the relevant part again where necessary. Keep track of changes affecting comparability.
The pilot is therefore a diagnostic step. It can improve the main investigation without pretending that a small preliminary sample proves the final hypothesis. A clear evaluation states the problem found and the mechanism of the improvement rather than simply add pilot study to a list of impressive procedures.
CHAPTER 14 OF 31 · Diagnose collection problems
14. Sample coverage determines which conditions are represented
A study’s geographical and time coverage affects how broadly its findings can be applied. Many observations from one narrow setting may still leave relevant conditions absent.
An invented park-use survey conducted beside the playground may capture particular users more often than users of a distant walking route.
Revising collection locations can address that coverage gap. Adding more people at the same spot may not.
State the intended population and the actual sample. The original evidence can still describe the sampled users. The limitation concerns the wider claim. This distinction allows a useful conclusion and a practical next step, rather than treating every incomplete sample as if it had collected no information at all.
CHAPTER 15 OF 31 · Connect samples and conclusions
15. More observations help when they address the uncertainty
A larger sample can provide more information, but its value depends on design. More readings do not automatically remove bias, clarify a vague question or capture missing locations.
Identify what additional observations would contribute. Repeating across days addresses temporal variation; adding suitable sites addresses spatial coverage; repeated instrument readings can examine short-term consistency.
These are different forms of more. Explain which one fits the investigation.
A strong answer can recommend a larger sample while making the reason explicit. For example, collecting at matched periods across several days helps assess whether a one-day pattern persists. That is more informative than the unsupported claim that doubling the sample makes the study twice as valid. Quality depends on the relationship between evidence and question.
CHAPTER 16 OF 31 · Connect samples and conclusions
16. Non-response can change whose views are recorded
People who decline or cannot participate may differ from those who answer. The direction of any resulting bias needs investigation; it should not be invented automatically.
Record relevant response counts under the approved method and keep skipped items distinct from neutral answers. A blank response does not establish a particular view.
If a study reaches people only through one channel or time, consider who might be absent. The issue connects recruitment to coverage.
An improvement can adjust the approach where appropriate, while the conclusion remains clear about the actual respondents. Do not claim that every non-response supports or opposes the hypothesis. Honest evaluation identifies the possible limitation and the evidence needed to understand it.
CHAPTER 17 OF 31 · Connect samples and conclusions
17. Repeated measurements need comparable conditions
Repetition is useful when its purpose is clear. Repeating under the same controlled conditions tests a different question from observing a changing place over several weeks.
If a weather variable changes naturally, different readings can reveal real variation. Averaging them may summarise that period without showing the cause.
Record conditions and choose an interval suited to the aim. A study of peak crowding may need a different design from one of typical daily use.
In evaluation, explain what repetition improves and what it leaves unresolved. It can reveal instability, reduce the influence of some random fluctuations or broaden temporal evidence, depending on the design. The word repeated alone does not establish which benefit occurred.
CHAPTER 18 OF 31 · Connect samples and conclusions
18. An average summarises evidence without repairing every weakness
A mean can reduce a set of values to one summary. It does not remove a systematic error or make incomparable observations equivalent.
An invented mean of morning and afternoon temperatures may describe that selected pair, while answering a noon comparison poorly. The arithmetic can be correct for an unsuitable summary.
Use the statistic that matches the question and retain relevant variation. A range or individual observations may be important when the aim concerns changing conditions.
The evaluation should connect the summary to the intended inference. Better presentation can make evidence clearer, but it cannot create coverage or a valid measure after the fact. This is why analysis choices belong in the design from the beginning.
CHAPTER 19 OF 31 · Connect samples and conclusions
19. Different methods can answer connected parts of a question
Combining methods can strengthen understanding when each has a clear role. A pedestrian count describes movement; a questionnaire can explore users’ reported experience; a map locates the setting.
The methods need not produce identical numbers because they measure different things. Agreement and difference should be interpreted through their purposes.
An invented busy route with low satisfaction can be a meaningful finding rather than a contradiction. People may use it because it is necessary.
Explain what each source contributes and how the pieces connect. Avoid saying multiple methods prove the conclusion merely because there are more sources. Their relevance, quality and relationship still matter. A thoughtful combination can reveal a better explanation than one measure alone.
CHAPTER 20 OF 31 · Connect samples and conclusions
20. A reliable pattern does not automatically prove cause
Even a consistent association can have several explanations. Another factor may influence both variables, or the direction of influence may differ from the proposed account.
An invented relationship between visit frequency and satisfaction does not establish that increasing visits will cause satisfaction. Satisfied users may choose to visit more.
Describe the pattern and then assess the mechanism using appropriate knowledge and additional evidence.
The correlation guide develops this distinction. In fieldwork evaluation, the important point is that consistency supports one aspect of the evidence while causal validity requires another line of reasoning. Keeping these claims separate makes the final conclusion stronger.
CHAPTER 21 OF 31 · Connect samples and conclusions
21. A conclusion can be useful within a narrower boundary
A finding may be well supported for the sampled sites and times while not representing the whole town. State that boundary rather than exaggerate or discard the result.
An invented survey of 20 park users can describe those responses. Wider resident preferences require an appropriate sampling basis and further evidence.
A useful conclusion identifies the sample, finding and limitation in connected sentences. It can then suggest a focused extension.
This is a positive way to evaluate small school investigations. Their value often lies in learning how to collect and interpret evidence carefully. A modest supported conclusion demonstrates more understanding than a sweeping claim whose scope cannot be justified.
CHAPTER 22 OF 31 · Choose targeted improvements
22. Worked task: choose an improvement for a timing problem
An invented study compares air temperature at two sites, one measured at 8 am and the other at 2 pm. The hypothesis concerns location differences.
The limitation is that the comparison mixes place with time. Changing daily conditions may contribute to the observed difference.
A suitable improvement is matched or simultaneous measurement where feasible, using comparable procedures. Repetition across suitable days can then test persistence.
A response recommending only a larger questionnaire sample would miss the actual issue. The task concerns temperature collection and comparability. State why the chosen revision helps and keep its limits clear: it reduces one confounding influence but does not automatically control every difference between sites. This is the full reasoning expected from an improvement.
CHAPTER 23 OF 31 · Choose targeted improvements
23. Worked task: repair a mismatch between aim and measure
An invented investigation asks whether a garden has greater biodiversity but collects only ratings of how attractive visitors find it. The ratings can support an appearance-perception finding.
They do not directly measure the variety of living organisms. The key weakness is the measure’s relationship to the aim.
One route is to add a suitable teacher-approved ecological method. Another is to narrow the research question to visitor perceptions. Explain which route the task requires.
Repeating the attractiveness question many times would not repair the original biodiversity inference. This example separates consistency from relevance. It shows how a clear diagnosis can produce a practical improvement without dismissing the value of the information already collected for a different question.
CHAPTER 24 OF 31 · Choose targeted improvements
24. Worked task: interpret a short repeated-count series
Use invented five-minute counts of 18, 20 and 19 under comparable stated conditions. Their mean is 19, with a range of two. This shows limited variation in that small set.
Now compare 5, 20 and 32. The mean is also 19, but the range is 27. The same average conceals different variation.
The difference may reflect real changes, collection inconsistency or both. Field notes and method information are needed before assigning a cause.
A strong answer describes the summaries and the uncertainty. It does not automatically delete the low or high value. The next step is to investigate the observations and decide what collection or analysis improvement addresses the actual issue. The arithmetic helps diagnose a question rather than settle every explanation.
| Invented series | Values | Mean | Range |
|---|---|---|---|
| A | 18, 20, 19 | 19 | 2 |
| B | 5, 20, 32 | 19 | 27 |
CHAPTER 25 OF 31 · Choose targeted improvements
25. Write evaluation as problem, consequence and improvement
Choose one specific limitation. Explain how it affects the finding, then propose a revision that changes that mechanism. Finally, state what the revision would and would not establish.
This sequence produces a reasoned paragraph rather than a list. It also makes it easier to see when a proposed improvement belongs to a different problem.
A parent or tutor can ask why the change helps. If the answer only repeats more reliable, the mechanism is still missing.
Practise with a new example after a gap. Move from timing to wording or from sample coverage to recording rules. Successful transfer shows the evaluation skill is understood. The aim is an independent judgement about evidence, not a memorised collection of impressive-sounding improvements.
CHAPTER 26 OF 31 · Choose targeted improvements
26. An unusual reading needs a reasoned decision
An observation far from the others can indicate a recording error, an unusual real condition or a different process. Its position alone does not identify which explanation applies.
Trace it to the original record and notes. Check time, place, units and the collection procedure before deciding whether correction or exclusion is justified.
Removing a point because it weakens the hypothesis can make the result look tidier while damaging the evidence.
A useful evaluation records the decision and its basis. If the value is valid, explain what it suggests about variation or an additional influence. The improvement may be better contextual recording rather than automatic deletion. This keeps the investigation responsive to evidence instead of training it to reproduce the expected pattern.
CHAPTER 27 OF 31 · Choose targeted improvements
27. Instrument resolution limits the differences you can interpret
An instrument displaying whole units does not support the same fine comparison as one recording tenths under an appropriate method. Read the resolution and instructions before interpreting very small differences.
An invented difference of a fraction of a unit may not be distinguishable from the equipment’s recording limits. Extra decimal places added during calculation do not create new observational precision.
Use suitable equipment when the question requires finer distinctions, while considering calibration and method.
The conclusion can state that the available readings do not clearly resolve a small difference. That is a meaningful finding about the evidence. It is better than asserting exact equality or a precise ranking that the measurement system cannot support.
CHAPTER 28 OF 31 · Choose targeted improvements
28. An unexpected result can still be a successful investigation
A study does not fail simply because its findings differ from the hypothesis. A hypothesis is a proposition to investigate, not an answer the data must be made to confirm.
First assess the method and evidence. Then decide whether the findings support, challenge or leave the proposition unresolved within the study’s scope.
An invented result showing no clear site difference can be informative if the collection was appropriate. It may suggest a narrower effect, more variation or another question.
A strong evaluation distinguishes a methodological weakness from an inconvenient outcome. Improving the method means strengthening the test, not guaranteeing the expected result. This is an optimistic learning principle: careful inquiry produces understanding through both confirmation and revision.
CHAPTER 29 OF 31 · Practise a reasoned evaluation
29. Combined percentages can conceal a sample imbalance
Imagine an invented survey with 18 respondents from one entrance and two from another. A combined percentage is dominated by the first group.
The arithmetic may be correct, while the interpretation of all park users remains limited by the sample’s distribution. Report relevant group counts and consider the aim.
Adding two more respondents at the first entrance does not necessarily address the missing coverage at the second.
A targeted improvement revises the sampling design in a way appropriate to the research question. Do not invent weights after collection without a justified method. The useful evaluation identifies how the distribution affects the inference and what evidence would strengthen a wider claim.
CHAPTER 30 OF 31 · Practise a reasoned evaluation
30. Method changes need to be documented
If a team changes the observation interval, questionnaire wording or site boundary during collection, later values may not be directly comparable with earlier ones.
Record what changed and why. The revision may be sensible, but its effect on the dataset still needs to be considered.
An invented switch from counting entries to counting everyone passing creates a different observation rule. Combining the totals silently would obscure that difference.
A clear report can separate versions or limit the analysis to comparable observations, according to the task and teacher’s guidance. Documentation preserves the link between numbers and the method that produced them. It also makes the proposed improvement reviewable rather than leaving future readers to assume the procedure remained unchanged.
CHAPTER 31 OF 31 · Practise a reasoned evaluation
31. Questions students and parents often ask
Are reliability and validity identical? They address different aspects of evidence and its use.
Does repetition always fix a study? It helps with specific uncertainties and can reproduce systematic problems.
Does variation prove unreliability? Conditions may genuinely change. Examine the method and context.
Can a consistent result answer the wrong question? Yes. The measure must match the aim.
Does a larger sample guarantee validity? No. Wording, coverage and inference still matter.
Can a small study be useful? Yes, with a clear supported scope.
What makes an improvement strong? Explain the problem it addresses and how the revised method strengthens the relevant evidence.
Continue with the relevant eduKate guide
Fieldwork and sampling: Design coverage suited to the question.
Questionnaire design: Repair wording and response-option problems.
Correlation: Separate consistent association from causation.
Unfamiliar evidence: Keep the claim within the measure and source.
Compare Geography learning routes. Use the guide matching your subject and current task; a G3 extension does not imply identical G2 assessment requirements.
Sources and next steps
SEAB: 2027 G2 Geography K226, Geographical Methods and assessment objectives · UK Government Analysis Function: questionnaire design and testing. The investigations and numerical datasets are original learning illustrations.
Worked examples and invented figures are learning illustrations. Check your school’s instructions and the official syllabus for your examination year.
Bring one current task and its feedback when seeking help. Ask eduKate about appropriate support and current availability. The Clementi Mathematics guide illustrates finding a learning gap before adding practice. Check Geography support separately.